AppWorld: A Controllable World of Apps and People — A New Testbed for Interactive Coding Agents

Why everyday digital chores are much harder than they look

AppWorld is a new Benchmark on for Coding LLMs. Image created with DALL-E.

When people talk about “AI assistants,” they often imagine a helpful bot that can answer emails, buy a gift, add the receipt to a folder, split the bill with roommates, and message the group chat — seamlessly. The hard part isn’t clicking a few buttons. It’s orchestrating many app…